An integrated intelligent operation system for comprehensive transportation hubs based on multi-source heterogeneous data

Abstract

The present invention discloses an integrated intelligent operation system for a comprehensive transportation hub based on multi-source heterogeneous data, which relates to the technical field of transportation control. The system comprises: a multi-source data processing unit, a data integration and analysis unit and a transportation hub operation management unit; the multi-source data processing unit is used to collect passenger flow data and transportation capacity data, and is also used to perform data cleaning and data storage on the collected data; the data integration and analysis unit is used to perform data integration and data feature extraction processing on the collected data to obtain unified representation information; the transportation hub operation management unit is used to predict the passenger flow dynamics of the comprehensive transportation hub, and to perform online evaluation of the hub operation status according to the prediction results, and to perform hub distribution and dispatching based on the hub operation status evaluation results and the hub comprehensive information. The present invention realizes accurate passenger flow prediction and efficient coordination and flexible dispatching of multi-mode and multi-organization transportation through multi-source data integration and analysis, thereby improving passengers' travel satisfaction.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic control and management, and in particular to an integrated intelligent operation system for a comprehensive transportation hub based on multi-source heterogeneous data. Background Art

[0002] With the rapid economic development and the acceleration of urbanization, the single transportation mode has gradually changed to intensive and integrated, forming a large-scale urban integrated transportation hub, which serves as a carrier for travelers to "seamlessly connect" and transfer, and highly gathers multiple modes of transportation in space. The intelligent, information-based and coordinated integrated transportation hub operation and management system can realize the safe and efficient operation of the hub and provide passengers with fast and orderly travel services.

[0003] The diverse passenger collection and distribution methods often make comprehensive transportation hubs prone to passenger congestion and tight transportation capacity resources in daily operation. The traditional transportation hub monitoring system only collects and analyzes single passenger flow information, lacks analysis and mining of hub vehicle flow, activity information, public area congestion and other data, and cannot fully and accurately grasp the internal operation status of the hub, and cannot make timely and effective judgments on future passenger flow, vehicle operation and other conditions. The lack of data integration and processing capabilities makes the operation and management efficiency low. Moreover, it cannot solve and optimize the problems of multiple units in comprehensive transportation hubs, weak coordination and linkage, inaccurate capacity scheduling, and inability to coordinate and dispatch overall. Summary of the invention

[0004] In view of this, the present invention provides an integrated intelligent operation system for a comprehensive transportation hub based on multi-source heterogeneous data. By constructing a comprehensive transportation hub database, effective screening and unified representation of data are achieved, and an integrated operation and management mode of a multi-mode comprehensive transportation hub is formed. The system solves the technical problems that the existing hub operation and management system lacks data integration and processing capabilities, has low operation and management efficiency, and cannot solve and optimize the problems of many units in the comprehensive transportation hub, weak coordination and linkage, inaccurate capacity scheduling, and inability to coordinate and coordinate scheduling globally.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The present invention proposes an integrated intelligent operation system for a comprehensive transportation hub based on multi-source heterogeneous data, including: a multi-source data processing unit, a data integration and analysis unit, and a transportation hub operation management unit;

[0007] The multi-source data processing unit is used to collect passenger flow data and transport capacity data, and perform data cleaning and data storage on the collected data;

[0008] The data integration analysis unit is used to perform data integration and data feature extraction processing on the data processed by the multi-source data processing unit to obtain unified representation information;

[0009] The transportation hub operation management unit is used to realize real-time passenger flow detection of the comprehensive transportation hub, and to perform dynamic passenger flow prediction based on unified characterization information, to conduct online evaluation of the hub operation status based on monitoring and prediction results, to obtain the optimal scheduling strategy and to perform scheduling of the comprehensive transportation hub.

[0010] Furthermore, the multi-source data processing unit includes a multi-source data acquisition module, a data cleaning module and a database module; the passenger flow data and capacity data include personnel and license plate information entering and leaving the hub, passenger flow information within the hub, and capacity information of various modes of transportation within the hub; the data cleaning module is used to clean the data collected by the multi-source data acquisition module; and the database module is used to store the data cleaned by the data cleaning module.

[0011] Furthermore, the data integration and analysis unit includes a data integration module and a data feature extraction module; the data integration module is used to integrate and connect various types of data stored in multi-source data processing units; the data feature extraction module is used to extract features of travel modes, travel time periods, travel groups, transfer behaviors, and activity types based on the integrated data to obtain unified representation information.

[0012] Furthermore, the transportation hub operation management unit includes a labeling processing module, a passenger flow dynamic prediction module, an operation status evaluation module, a real-time passenger flow monitoring module, a scheduling analysis module and a flexible scheduling module;

[0013] The labeling processing module is used to label the unified representation information extracted by the data feature extraction module. The labeling processing uses predefined labels including weather influencing factors, week influencing factors and non-fixed attribute influencing factors for marking. The unified representation information after labeling is used to extract multi-granularity features of passenger flow in time and space. The multi-granularity features of passenger flow in time and space include station entry and exit volume features, section passenger flow features, line passenger flow features, and starting and ending point features.

[0014] The passenger flow dynamic prediction module includes a transfer passenger flow distribution analysis module, a regular passenger flow prediction module and an abnormal passenger flow prediction module. The transfer passenger flow distribution analysis module is used to establish a hub transfer prediction model to predict the transfer passenger flow between various modes of transportation and analyze the transfer passenger flow distribution. The regular passenger flow prediction module is used to establish a short-term passenger flow prediction model using the spatiotemporal multi-granularity characteristics of passenger flow under normal conditions to predict regular passenger flow within the hub. The abnormal passenger flow prediction module is used to establish an abnormal passenger flow prediction model based on the spatiotemporal multi-granularity characteristics of passenger flow under abnormal conditions to predict abnormal passenger flow within the hub.

[0015] The operation status assessment module is used to assess the capacity of the transportation hub and identify the traffic operation status using the AHP model based on the regular passenger flow prediction data, the abnormal passenger flow prediction data, the predicted transfer passenger flow and the number of people and vehicles in the hub;

[0016] The real-time passenger flow monitoring module is used to further analyze the actual real-time passenger flow and average passenger stay time of the hub according to the real-time passenger flow information collected by the multi-source data collection module, and output it as real-time passenger flow monitoring information;

[0017] The dispatch analysis module is used to solve the optimal dispatch strategy based on real-time passenger flow monitoring information, passenger flow dynamic prediction results and operation status evaluation results, so as to minimize the comprehensive passenger transfer time, passenger waiting time and hub traffic dispatch operation cost;

[0018] The flexible scheduling module is used to schedule the comprehensive transportation hub according to the optimal scheduling strategy.

[0019] Furthermore, the establishment of a hub transfer prediction model to predict the transfer passenger flow between various transportation modes includes:

[0020] The hub transfer prediction model is constructed based on the gravity model and trained using the spatiotemporal multi-granularity features of passenger flow. The trained hub transfer prediction model is used to predict the transfer passenger flow between various modes of transportation. The gravity model is:

[0021]

[0022] Among them, q ij is the predicted value of the transfer passenger flow between hub site i and hub site j; Q i and D j are the number of transfer passengers generated by hub station i and the number of transfer passengers attracted by hub station j, respectively, which are obtained based on the spatiotemporal multi-granularity feature extraction of passenger flow; c ij is the transfer time between hub station i and hub station j; α and β are model parameters.

[0023] Furthermore, the short-term passenger flow prediction model is established by using the spatiotemporal multi-granularity characteristics of passenger flow under normal conditions to predict the regular passenger flow in the hub, specifically including: classifying the spatiotemporal multi-granularity characteristics of passenger flow according to the clustering algorithm, using different categories of spatiotemporal multi-granularity characteristics of passenger flow to establish and train the SVM model, the trained SVM model is the short-term passenger flow prediction model, and the short-term passenger flow prediction model is used to perform nonlinear regression prediction on the outbound passenger flow of each station to obtain regular passenger flow prediction data. ;

[0024] Furthermore, the abnormal passenger flow prediction model is established to predict abnormal passenger flow within the hub, including:

[0025] Constructing an abnormal passenger flow prediction model, wherein the abnormal passenger flow prediction model includes an elastic network and a multi-dimensional feature recognition model;

[0026] Obtain information about abnormal events that occur in the hub, filter out the spatiotemporal multi-granularity features of passenger flow during abnormal events, and extract multi-dimensional sudden large passenger flow features based on the elastic network. The formula of the elastic network is:

[0027] y=β0+β1x1+β2x2+…β p x p

[0028] Among them, y is the target hub passenger flow sequence to be predicted, that is, the multi-dimensional sudden large passenger flow characteristics, x1,x2,…,x p is the spatiotemporal multi-granularity feature of passenger flow, p is the number of spatiotemporal multi-granularity feature variables of passenger flow, β1,β2,…,β p are model parameters, which are optimized based on the objective function G(β):

[0029]

[0030] in, β j The L1 norm of β j The L2 norm of n is the number of hub sites, y i is the predicted target hub passenger flow of the ith station, x ij is the passenger flow characteristics of the i-th station;

[0031] The extracted multi-dimensional sudden passenger flow features and historical passenger flow spatiotemporal multi-granularity features are input into the multi-dimensional feature recognition model, and the passenger flow feature vector after disturbance removal is calculated according to the following formula:

[0032]

[0033] Among them, y t-1 ,…,y t-n Represents the spatiotemporal multi-granularity characteristics of passenger flow at different times in history, represents the disturbance of passenger flow and abnormal events at other hubs to the target hub, represents the multi-dimensional sudden passenger flow characteristics at the prediction time, i t represents the observed value at time point t; Represents the historical passenger flow feature vector without disturbance; the output of the multi-dimensional feature recognition model is combined with the disturbance at the prediction moment to obtain the final abnormal passenger flow prediction result.

[0034] Furthermore, the further analysis of the actual real-time passenger flow and average passenger stay time of the hub based on the real-time passenger flow information collected by the multi-source data collection module specifically includes:

[0035] The real-time passenger flow monitoring module obtains the actual real-time passenger flow of the hub based on the real-time passenger flow information collected by the multi-source data collection module and by combining the mobile phone signaling data of the hub staff:

[0036]

[0037] Where T is the total instantaneous passenger flow at the hub. The multi-source data collection module is equipped with n base stations in the hub for collecting real-time passenger flow information. i is the number of people at the i-th base station, ∑C a is the number of staff members of the hub, τ is the conversion coefficient, and the instantaneous total passenger flow in the hub during the historical period is input into the time series model to predict the future short-term passenger flow in the hub;

[0038] The real-time passenger flow monitoring module also analyzes the average passenger stay time through the following formula:

[0039]

[0040] in, represents the residence time of the u-th passenger at the i-th base station, represents the average stay time of N passengers at the hub.

[0041] Further, the solving of the optimal dispatching strategy based on the real-time passenger flow monitoring information, the passenger flow dynamic prediction results and the operation status evaluation results includes: the dispatching analysis module determines the passenger flow, vehicle flow, congestion status of public areas and vehicle demand prediction information inside and outside the hub based on the real-time passenger flow monitoring information, the passenger flow dynamic prediction results and the hub operation status evaluation results, and establishes an objective function that minimizes the comprehensive passenger transfer time, passenger waiting time and hub traffic dispatching operation cost, the solution of the function is a set of departure time sets with unequal intervals, and a genetic algorithm is used to solve it, and the optimal dispatching strategy is obtained according to the solution result;

[0042] The dispatching of the comprehensive transportation hub according to the optimal dispatching strategy includes: the flexible dispatching module adjusts the traffic mode and departure interval by adjusting the bus schedule, taxi call, and subway entrance elevator operation according to the optimal dispatching strategy; it also adjusts the parking position, guides the passenger flow, and publishes the passenger flow situation to the vehicle terminal.

[0043] It can be seen from the above technical solution that the advantages of the present invention are:

[0044] 1. This application comprehensively and accurately grasps the internal operation status of the hub through comprehensive analysis and mining of multi-source data such as passenger flow information, hub vehicle flow, activity information, and congestion conditions in public areas, so as to make timely and effective judgments on real-time passenger flow, future passenger flow, vehicle operation, etc., and improve data processing efficiency through data integration processing methods, making operation management more efficient.

[0045] 2. This application also uses the passenger flow and hub situation forecast data for conventional and special scenarios of integrated transportation hubs to determine the statistics of passengers and vehicles inside and outside the hub, and selects the optimal control plan through comprehensive multi-dimensional passenger models and indicators, so as to achieve multi-department, multi-regional linkage management, multi-mode, and multi-organizational efficient coordination and flexible scheduling of transportation. Through the predicted passenger flow trends combined with the management plan formulated between multiple organizational departments, passengers are provided with end-to-end services, and various types of transportation are reasonably scheduled, so as to achieve efficient passenger diversion and control in multi-mode transportation hubs in conventional or special scenarios, reduce urban traffic congestion and waste of resources, and improve passenger travel satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0047] Figure 1 It is a structural schematic diagram of the present invention.

[0048] Figure 2 It is a detailed structural schematic diagram of the present invention.

[0049] Figure 3 It is a schematic diagram of the process of abnormal passenger flow prediction of the present invention.

[0050] Figure 4 It is a structural schematic diagram of the transportation hub operation management unit of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0052] The comprehensive transportation hub integrates multiple transportation modes such as railways, long-distance passenger transport, subways, buses and taxis, and is an important node in the urban transportation system. The integrated operation system of the comprehensive transportation hub of the present invention is committed to achieving seamless connection and efficient operation between various transportation modes.

[0053] like Figure 1 and Figure 2As shown, this embodiment provides an integrated intelligent operation system for a comprehensive transportation hub based on multi-source heterogeneous data. Through the collaborative work of multiple modules and combined with multi-source data processing technology, efficient hub operation and intelligent management are achieved, and the travel experience of passengers is improved. The system specifically includes: a multi-source data processing unit, a data integration and analysis unit, and a transportation hub operation management unit. The multi-source data processing unit collects multi-source data in different ways, and performs operations such as cleaning on the collected data to improve the data quality. The data integration and analysis unit performs integrated association, feature selection and extraction to ensure that the data is suitable for subsequent data analysis or model training. The transportation hub operation management unit provides decision support for operation management by real-time monitoring and analysis of the passenger flow and capacity data of the hub.

[0054] Specifically, the multi-source data processing unit is used to collect passenger flow data and capacity-related data, and to store the collected data and system data, and is also used to clean and store the collected data.

[0055] The multi-source data processing unit includes a multi-source data acquisition module, a data cleaning module and a database module; the multi-source data acquisition module includes an image acquisition network, an on-board GPS, a traffic Internet of Things, a sensor network and an indoor positioning module. The image acquisition network collects the information of people entering and leaving the hub and the license plate information of vehicles entering and leaving the hub through cameras. The on-board GPS and the traffic Internet of Things are used to collect and count vehicle import and export data, obtain railway data, subway data, bus data, aircraft data and road conditions, the sensor network is used to collect environmental information, and the indoor positioning module is used to collect passenger flow information in the hub using a location positioning method that combines LampSite and SVA. The data cleaning module is used to clean the time series data corresponding to historical passenger flow data, railway data, subway data, bus data and airplane data, as well as the comprehensive transportation hub all-round video image data and real-time passenger flow information acquired in real time by the equipment of the multi-source data acquisition module, in combination with the inherent attributes of the data, to remove duplicate, erroneous or inconsistent data, ensure the quality and accuracy of the data, and store the cleaned data; data cleaning methods such as clustering algorithms and sorting neighbor methods for processing structured data, and adaptive redundant image screening algorithms for processing unstructured data, including operations such as deduplication, removal of null values, data type conversion, and data format conversion. This embodiment collects information about people and vehicles in real time through mobile phone networks, global positioning system (GPS) / Beidou vehicle navigation, and transportation Internet of Things.

[0056] like Figure 2As shown, the database module in the multi-source data processing unit is used to store the structured and unstructured data obtained after cleaning; the operation of the integrated transportation hub involves the support of multiple databases, including basic database, business database, exchange database and subject database. The basic database stores the basic information and data of the hub, such as geographical location, building structure, equipment and facilities, and personnel information, including the hub's CAD drawings, BIM models, equipment and facilities lists, and personnel information tables; the business database is used to store and manage the daily operation business data of the integrated transportation hub, such as ticketing data, passenger flow data, and vehicle operation data, including various business forms, statistical reports, and data analysis results; the exchange database is mainly used to realize data exchange and sharing between different systems in the integrated transportation hub, and ensure the circulation of data between different systems by formulating unified data exchange standards and interface specifications; the subject database is used to store and manage data related to the corresponding subject or business needs, including various data tables, views, and indexes related to the corresponding subject or business needs. For example, the security monitoring database includes security monitoring, emergency response, accident records and security-related data.

[0057] The data integration and analysis unit is connected to the multi-source data processing unit, and is used to perform data integration and data feature extraction processing on the collected data to obtain unified representation information.

[0058] The data integration and analysis unit includes a data integration module and a data feature extraction module; the data integration module is used to integrate historical passenger flow data, current passenger flow trends, capacity data of various transportation modes, and connect various types of data; the data feature extraction module is used to extract data features based on preprocessed and simply associated data using a feature extraction model, the preprocessed data is data that has undergone data cleaning processing, and the simply associated data includes data that has undergone connection processing.

[0059] When connecting various types of data, the data integration module first analyzes based on association rules. The association rule analysis process includes: relying on multi-source transportation big data, using Apriori and graph clustering algorithms to mine frequent item sets and conduct hub correlation analysis, generate interest indexes, and summarize association rules between multi-source data sets. Frequent item sets refer to items that often appear together, such as traffic modes, travel times, starting points, and destinations. Multi-source transportation big data includes multi-period hub inflow and outflow data, public transportation card swiping data, and residents' travel attributes, which are collected through multi-source data processing units. Subsequently, the data feature extraction module fuses the data sets according to the association rules, that is, using the association rules as a guide to determine which data sets should be fused to extract relevant features, and extracts features and sorts interest indicators for various travel modes, travel time periods, travel groups, transfer behaviors, and activity types, respectively, to achieve parameterization and structuring of the travel process within the transportation hub. Among them, the support, confidence, and lift metrics calculated in the association rules are used to sort and evaluate the extracted features. The support indicates the frequency of the feature in the data set, the confidence indicates the probability of another feature appearing when one feature appears, and the lift measures the correlation or independence between the features. Based on these indicators, the most important features or feature combinations can be identified to provide support for subsequent model training and prediction.

[0060] like Figure 2 As shown in the figure, data feature extraction can extract targeted information from integrated multi-source heterogeneous data to obtain unified representation information. The data feature extraction module uses the feature extraction model to extract representative features from the preprocessed data and converts them into a data form that can be processed by subsequent algorithms, thereby improving the accuracy and efficiency of data analysis and machine learning.

[0061] Based on the data analysis results, efficient management of transportation hub operation can be achieved. The transportation hub operation management unit is used to predict the passenger flow dynamics of the comprehensive transportation hub, and conduct online evaluation of the hub operation status based on the prediction results, and perform hub distribution dispatch based on the hub operation status evaluation results and hub comprehensive information.

[0062] like Figure 4As shown in FIG, the transportation hub operation management unit includes a labeling processing module, a passenger flow dynamic prediction module and an operation status evaluation module; the labeling processing module is used to label the extracted unified representation information, and the unified representation information after labeling is used to extract the spatiotemporal multi-granularity features of passenger flow to provide input features for the passenger flow dynamic prediction module. The spatiotemporal multi-granularity features of passenger flow include station entry and exit volume features, section passenger flow features, line passenger flow features, and OD (origin and destination) features. The processed spatiotemporal multi-granularity features of passenger flow are used as the input of the passenger flow dynamic prediction module; the passenger flow dynamic prediction module is used to establish a hub transfer prediction model to analyze the transfer passenger flow distribution, and the transfer passenger flow distribution is analyzed. The analysis includes analyzing the expected number of passengers arriving and departing at the hub by each connection mode, the number of passengers shared by each mode of transportation, and the proportion of passenger flow characteristics, and using the real-time travel characteristics of passengers under normal circumstances and the historical passenger flow characteristics to establish a short-term passenger flow prediction model based on a time series model to predict regular passenger flow in the hub, and according to the historical passenger flow information under abnormal circumstances, establish an abnormal passenger flow prediction model to predict abnormal hub passenger flow. The hub transfer prediction model, short-term passenger flow prediction model and abnormal passenger flow prediction model are constructed using the SVM model structure; the operation status evaluation module is used to construct an online evaluation model of the operation status of an urban comprehensive transportation hub based on the passenger flow prediction data, the transfer passenger flow distribution information and the number of vehicles and passengers counted by the hub using the AHP (Analytic Hierarchy Process) model, evaluate the capacity of the transportation hub, and identify the traffic operation status.

[0063] Specifically, the labeling process uses predefined labels including weather influencing factors, week influencing factors, and non-fixed attribute influencing factors to classify the unified representation information, thereby making data analysis more accurate and goal-oriented.

[0064] In this embodiment, the weather influencing factors include temperature, humidity, wind speed, and precipitation; the weekly influencing factors include the weekly identifier; the non-fixed attribute influencing factors include special events such as holidays, concerts, and sports events. Passenger flow data includes station in and out volume, section passenger flow, line passenger flow, and OD (origin and destination) passenger flow. The abnormal situation refers to major events or emergencies such as large-scale performances, extreme weather, epidemic infectious diseases, and natural disasters in the hub. The traffic operation status includes unimpeded state, slow state, congested state, stagnant state, and changing state (vehicle speed and flow are unstable and change frequently). .

[0065] Specifically, the passenger flow dynamic prediction module includes a regular passenger flow prediction module, a transfer passenger flow distribution analysis module and an abnormal passenger flow prediction module; the regular passenger flow prediction module is used to classify the passenger flow according to the clustering algorithm, and the passenger flow time series of different categories are modeled using the time series model respectively, and the SVM model is trained to obtain a short-term passenger flow prediction model. The short-term passenger flow prediction model is used to perform nonlinear regression prediction on the station exit passenger flow to obtain passenger flow prediction data. When training the SVM model, it is necessary to determine the kernel function and parameters. In this embodiment, the RBF kernel function is used, and the firefly algorithm is used to optimize the parameters of the RBF kernel function to obtain its optimal parameters; the transfer passenger flow distribution analysis module is used to analyze the transfer passenger flow distribution according to the transfer prediction model; the abnormal passenger flow prediction module is used to predict the station exit passenger flow under large-scale activities and emergencies based on the abnormal passenger flow prediction model.

[0066] The hub transfer passenger flow is the passenger flow exchanged between different modes of transportation within the hub. Specifically, the analysis of the transfer passenger flow distribution according to the hub transfer prediction model includes:

[0067] The hub transfer prediction model is a trained gravity model. Based on the total passenger flow prediction value, the gravity model q ij Estimate the transfer passenger flow between various modes of transportation, the gravity model is: Among them, q ij is the predicted value of the transfer passenger flow between hub site i and hub site j; Q i and D j are the number of transfer passengers generated by hub site i and the number of transfer passengers attracted by hub site j respectively; c ij is the transfer time between hub site i and hub site j; α and β are model parameters, which can be calculated based on the travel distribution matrix and impedance matrix. The travel distribution matrix is ​​the geographical spatial relationship of each hub, and the impedance matrix is ​​the transfer time between different transportation modes.

[0068] The existing prediction model can predict the total passenger flow in the hub station, but lacks a method for predicting the transfer passenger flow between different modes of transportation. Therefore, this embodiment uses a gravity model to estimate the transfer passenger flow between various modes of transportation.

[0069] The characteristics of hub passenger flow mainly include time characteristics, spatial characteristics and activity characteristics. The passenger flow of hub stations presents the characteristics of continuous and dynamic fluctuations in the time series. Taking each time point as an independent variable unit, the passenger flow characteristics at different times can be understood as multiple one-dimensional vectors that change with time; spatial characteristics include both the passenger flow of different associated stations at the same time and the passenger flow of each associated station at different times, forming a spatiotemporal multidimensional vector; activity characteristics include large-scale activity category variables, influencing time period variables and passenger flow increments. In order to solve the problem of measuring the correlation between feature variables and prediction targets, an elastic network is used to analyze and extract passenger flow characteristics. Abnormal passenger flow is then predicted by a multidimensional feature recognition model, which includes a two-dimensional convolutional neural network based on a spatial attention model and a neural network based on a stacked autoencoder.

[0070] like Figure 3 Specifically, the prediction of the outbound passenger flow of a station under large-scale activities and emergencies based on the abnormal passenger flow prediction model includes:

[0071] Constructing an abnormal passenger flow prediction model, wherein the abnormal passenger flow prediction model includes an elastic network and a multi-dimensional feature recognition model;

[0072] Based on the elastic network, the multi-dimensional sudden large passenger flow characteristics are extracted to determine the passenger flow information of the peak period of passenger flow in and out of the comprehensive hub, the hub station association and the influence period of large-scale activities. The peak period of passenger flow in and out of the comprehensive hub is XX-XX, and the model formula is: y = β0 + β1x1 + β2x2 + ... β p x p , where the target hub passenger flow sequence to be predicted is y, the number of characteristic variables is p, and the passenger flow characteristics are (x1, x2,…, x p ), is the optimal regression coefficient {β1,β2,…,β p}, construct the objective function: in, β j The L1 norm of can simplify high-dimensional features. β j The L2 norm of can reduce the complexity of feature coefficients;

[0073] Since passenger flow includes both regular model passenger flow and irregular disturbance, disturbance needs to be processed when passenger flow prediction modeling is performed. The extracted multi-dimensional sudden large passenger flow characteristics and historical passenger flow characteristics data are input into the multi-dimensional feature recognition model, according to the formula: Calculate the historical passenger flow feature vector without disturbance, where y t-1 ,…,y t-nRepresents the spatiotemporal multi-granularity characteristics of passenger flow at different times in history, represents the disturbance of passenger flow and abnormal events at other hubs to the target hub, represents the multi-dimensional sudden passenger flow characteristics at the prediction time, i t represents the observed passenger flow at time point t, Represents the historical passenger flow feature vector without disturbance. Input the historical passenger flow feature vector without disturbance, and compare the output of the multi-dimensional feature recognition model with the disturbance at the prediction time. Combined with the above, the final abnormal passenger flow prediction result is obtained.

[0074] The transportation hub operation management unit also includes a real-time passenger flow monitoring module. The real-time passenger flow monitoring module is used to obtain real-time passenger flow based on the information collected by the indoor positioning module by combining mobile phone signaling data. It is also used to analyze the real-time passenger flow and gathering and distribution time of the hub. The analysis process is as follows:

[0075] Step S1: By formula: Calculate the passenger flow information at the station, where T is the total passenger flow at the hub at the instant, C i is the passenger flow at the i-th base station, n is the number of base stations, ∑C a is the number of staff members of the hub, τ is the conversion coefficient, and the passenger flow situation can be judged through statistical analysis based on the length of time each passenger stays at the base station. The total instantaneous passenger flow information of the hub and the historical passenger flow information are input into the time series model to predict the future short-term passenger flow in the hub;

[0076] Step S2: By formula: Calculate the average length of time passengers stay in the hub, and then analyze the hub's distribution, including: represents the residence time of the u-th passenger at the i-th base station, represents the average stay time of N passengers at the hub.

[0077] In this embodiment, the indoor positioning module uses LampSite and SVA technology to collect hub passenger flow data through multiple base stations, and then analyzes the passenger flow status and development trend, and the evaluation results of passenger flow operation capacity through bus, subway and taxi data, and pushes relevant information to surrounding taxis, accurately dispatches public transportation such as buses and subways, and avoids the occurrence of large-scale passenger stranded conditions. LampSite has a three-layer architecture of pRRU, RHUB, and BBU. RHUB is similar to a hub (or switch), and BBU is a baseband processing unit. The BBU baseband processing unit is first connected to RHUB. RHUB is similar to a hub and can connect multiple pRRUs. pRRU can be regarded as a "miniaturization" of RRU on a traditional base station, and the transceiver antenna is integrated inside the pRRU. The conversion between RF signals and baseband signals can be achieved through this three-layer architecture. SVA (Service Anchor) is a capability open platform based on a digital small cell network, which can provide location analysis and calculation functions and data transmission APIs based on three methods: field strength triangulation positioning, TDOA positioning, and fingerprint positioning. The accuracy of the SVA positioning algorithm can effectively distinguish passenger flows in the riding areas of each mode of transportation, as well as pedestrians and community residents around the hub.

[0078] In this embodiment, predicting the future short-term passenger flow in the hub specifically includes: establishing a corresponding time series model to predict passenger flow, and using the model to predict future values, specifically as follows: for a sequence X containing n observation results = {x1, x2, ..., x n}, predict the value of x(k+r) forward, then the linear or nonlinear regression process is: x(k+r)=f[x(k),x(k-1),x(k-2),…,x(kn-1), where r is the prediction step size and f is the relationship function between the past observations of the time series X and the current values.

[0079] In this embodiment, the trained SVM model is used to perform nonlinear regression prediction on the passenger flow to obtain passenger flow prediction data, and the firefly algorithm is used to optimize the parameters of the RBF kernel function of the SVM model to obtain its optimal parameters.

[0080] Understanding the real-time passenger flow status and predicting the future passenger flow scale are of great significance to achieve accurate hub capacity scheduling and avoid large-scale passenger stranding. When predicting passenger flow, first classify the passenger flow according to the clustering algorithm. For example, the passenger flow data of the day before the holiday is classified as Class A, and the passenger flow of holidays and weekends is similar, which is classified as Class B. The passenger flow time series of different categories are modeled using time series models. For example, the passenger flow of the first 6 time slices with high correlation is used to predict the passenger flow of future time slices. The calculation formula of the short-term passenger flow prediction model is as follows: x(k+1)=f[x(k),x(k-1),x(k-2),…,x(k-5). Use these 6 time slices for training and prediction to obtain future passenger flow prediction information.

[0081] In another embodiment of the present invention, the transportation hub operation management unit also includes a scheduling analysis module and a flexible scheduling module; the scheduling analysis module is used to determine the passenger flow, vehicle flow, congestion in public areas and vehicle demand prediction information inside and outside the hub based on the transfer passenger flow prediction information and distribution law between hubs, real-time passenger flow monitoring information, passenger flow dynamic prediction information and online evaluation results of the hub operation situation, and establish an objective function that minimizes the comprehensive passenger transfer time, regular passenger waiting time and urban traffic scheduling operation cost, the solution of the function is a set of departure time sets with unequal intervals, and a genetic algorithm is used to solve it. According to the solution result, an optimal scheduling strategy is obtained, and the organization and scheduling methods of various departments in the comprehensive transportation hub are optimized according to the optimal scheduling strategy to form an integrated scheduling plan library between hubs; the flexible scheduling module is connected to the scheduling analysis module, and the flexible scheduling module is used to coordinate various departments according to the optimal scheduling strategy by adjusting bus schedules, taxi calls, and subway entrance elevator operations, flexibly adjusting traffic modes and departure intervals, and providing real-time exit transfer options; and adjusting parking positions and guiding passenger flow, publishing passenger flow conditions to the on-board terminal, optimizing vehicle traffic routes and parking management.

[0082] As an important node of urban transportation, the comprehensive transportation hub gathers various modes of transportation such as road transportation, railway, rail transportation, bus and taxi, and generates a large amount of traffic data every day. These data contain rich passenger flow and traffic volume information, which are characterized by a wide variety, large temporal and spatial scale, dynamic and changeable, local, heterogeneous, short life cycle and high randomness. The present invention is based on the comprehensive transportation hub operation system of big data, which collects a large amount of hub-related traffic data by building a hub big data operation system, and uses big data technology to analyze and process data, integrates the analysis results of various types of traffic data, realizes passenger flow monitoring and hub information release functions, and achieves multi-department, multi-regional linkage management, multi-mode, multi-organization traffic efficient coordination and flexible scheduling.

[0083] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An integrated intelligent operation system for comprehensive transportation hubs based on multi-source heterogeneous data, characterized in that: include: Multi-source data processing unit, data integration and analysis unit, and transportation hub operation and management unit; The multi-source data processing unit is used to collect passenger flow data and transport capacity data, and perform data cleaning and data storage on the collected data; The data integration analysis unit is used to perform data integration and data feature extraction processing on the data processed by the multi-source data processing unit to obtain unified representation information; The transportation hub operation management unit is used to realize real-time passenger flow detection of the comprehensive transportation hub, and perform dynamic passenger flow prediction based on the unified representation information, and conduct online evaluation of the hub operation status based on the monitoring and prediction results, obtain the optimal scheduling strategy and perform scheduling of the comprehensive transportation hub; The multi-source data processing unit includes a multi-source data acquisition module and a data cleaning module; The passenger flow data and capacity data include the information of people and license plates entering and leaving the hub, the passenger flow information in the hub, and the capacity information of each mode of transportation in the hub; the data cleaning module is used to clean the data collected by the multi-source data collection module; The transportation hub operation management unit includes a labeling processing module, a passenger flow dynamic prediction module, an operation status evaluation module, a real-time passenger flow monitoring module, a scheduling analysis module and a flexible scheduling module; The labeling processing module is used to label the unified representation information extracted by the data feature extraction module. The labeling processing uses predefined labels including weather influencing factors, week influencing factors and non-fixed attribute influencing factors for marking. The unified representation information after labeling is used to extract multi-granularity features of passenger flow in time and space. The multi-granularity features of passenger flow in time and space include station entry and exit volume features, section passenger flow features, line passenger flow features, and starting and ending point features. The passenger flow dynamic prediction module includes a transfer passenger flow distribution analysis module, a regular passenger flow prediction module and an abnormal passenger flow prediction module. The transfer passenger flow distribution analysis module is used to establish a hub transfer prediction model to predict the transfer passenger flow between various modes of transportation and analyze the transfer passenger flow distribution. The regular passenger flow prediction module is used to establish a short-term passenger flow prediction model using the spatiotemporal multi-granularity characteristics of passenger flow under normal conditions to predict regular passenger flow within the hub. The abnormal passenger flow prediction module is used to establish an abnormal passenger flow prediction model based on the spatiotemporal multi-granularity characteristics of passenger flow under abnormal conditions to predict abnormal passenger flow within the hub. The operation status assessment module is used to assess the capacity of the transportation hub and identify the traffic operation status using the AHP model based on the regular passenger flow prediction data, the abnormal passenger flow prediction data, the predicted transfer passenger flow and the number of people and vehicles in the hub; The real-time passenger flow monitoring module is used to further analyze the actual real-time passenger flow and average passenger stay time of the hub according to the real-time passenger flow information collected by the multi-source data collection module, and output it as real-time passenger flow monitoring information; The dispatch analysis module is used to solve the optimal dispatch strategy based on real-time passenger flow monitoring information, passenger flow dynamic prediction results and operation status evaluation results, so as to minimize the comprehensive passenger transfer time, passenger waiting time and hub traffic dispatch operation cost; The flexible scheduling module is used to schedule the comprehensive transportation hub according to the optimal scheduling strategy; The method for solving the optimal dispatching strategy based on the real-time passenger flow monitoring information, the passenger flow dynamic prediction results and the operation status evaluation results includes: the dispatching analysis module determines the passenger flow, vehicle flow, congestion status of public areas and vehicle demand prediction information inside and outside the hub based on the real-time passenger flow monitoring information, the passenger flow dynamic prediction results and the hub operation status evaluation results, and establishes an objective function for minimizing the comprehensive passenger transfer time, passenger waiting time and hub traffic dispatching operation cost, wherein the solution of the function is a set of departure time sets with unequal intervals, and a genetic algorithm is used for solving the problem, and the optimal dispatching strategy is obtained according to the solution result; The dispatching of the comprehensive transportation hub according to the optimal dispatching strategy includes: the flexible dispatching module adjusts the traffic mode and departure interval by adjusting the bus schedule, taxi call, and subway entrance elevator operation according to the optimal dispatching strategy; it also adjusts the parking position, guides the passenger flow, and publishes the passenger flow situation to the vehicle terminal.

2. The integrated intelligent operation system for comprehensive transportation hubs based on multi-source heterogeneous data according to claim 1 is characterized in that: The multi-source data processing unit also includes a database module, and the database module is used to store the data cleaned by the data cleaning module.

3. The integrated intelligent operation system for comprehensive transportation hubs based on multi-source heterogeneous data according to claim 1 is characterized in that: The data integration and analysis unit includes a data integration module and a data feature extraction module; the data integration module is used to integrate and connect various types of data stored in multi-source data processing units; the data feature extraction module is used to extract features of travel modes, travel time periods, travel groups, transfer behaviors, and activity types based on the integrated data to obtain unified representation information.

4. The integrated intelligent operation system for comprehensive transportation hubs based on multi-source heterogeneous data according to claim 1 is characterized in that: The establishment of a hub transfer prediction model to predict the transfer passenger flow between various transportation modes includes: The hub transfer prediction model is constructed based on the gravity model and trained using the spatiotemporal multi-granularity features of passenger flow. The trained hub transfer prediction model is used to predict the transfer passenger flow between various modes of transportation. The gravity model is: Among them, q ij is the predicted value of the transfer passenger flow between hub site i and hub site j; Q i and D j are the number of transfer passengers generated by hub station i and the number of transfer passengers attracted by hub station j, respectively, which are obtained based on the spatiotemporal multi-granularity feature extraction of passenger flow; c ij is the transfer time between hub station i and hub station j; α and β are model parameters.

5. The integrated intelligent operation system for comprehensive transportation hubs based on multi-source heterogeneous data according to claim 1 is characterized in that: The method of establishing a short-term passenger flow prediction model by utilizing the spatiotemporal multi-granularity characteristics of passenger flow under normal circumstances to predict regular passenger flow within the hub specifically includes: classifying the spatiotemporal multi-granularity characteristics of passenger flow according to a clustering algorithm, establishing and training an SVM model using different categories of spatiotemporal multi-granularity characteristics of passenger flow, the trained SVM model is the short-term passenger flow prediction model, and utilizing the short-term passenger flow prediction model to perform nonlinear regression prediction on the outbound passenger flow of each station to obtain regular passenger flow prediction data.

6. The integrated intelligent operation system for comprehensive transportation hubs based on multi-source heterogeneous data according to claim 1 is characterized in that: Establishing an abnormal passenger flow prediction model to predict abnormal passenger flow within the hub specifically includes: Constructing an abnormal passenger flow prediction model, wherein the abnormal passenger flow prediction model includes an elastic network and a multi-dimensional feature recognition model; Obtain information about abnormal events that occur in the hub, screen the spatiotemporal multi-granularity features of passenger flow during abnormal events, and extract multi-dimensional sudden large passenger flow features based on an elastic network. The formula of the elastic network is: y=β0+β1x1+β2x2+…β p x p Among them, y is the target hub passenger flow sequence to be predicted, that is, the multi-dimensional sudden large passenger flow characteristics, x1,x2,…,x p is the spatiotemporal multi-granularity feature of passenger flow, p is the number of spatiotemporal multi-granularity feature variables of passenger flow, β1,β2,…,β p are model parameters, which are optimized based on the objective function G(β): in, β j The L1 norm of β j The L2 norm of n is the number of hub sites, y i is the predicted target hub passenger flow of the ith station, x ij is the passenger flow characteristics of the i-th station; The extracted multi-dimensional sudden passenger flow features and historical passenger flow spatiotemporal multi-granularity features are input into the multi-dimensional feature recognition model, and the passenger flow feature vector after disturbance removal is calculated according to the following formula: Among them, y t-1 ,…,y t-n Represents the spatiotemporal multi-granularity characteristics of passenger flow at different times in history, represents the disturbance of passenger flow and abnormal events at other hubs to the target hub, represents the multi-dimensional sudden passenger flow characteristics at the prediction time, i t represents the observed value at time point t; Represents the historical passenger flow feature vector without disturbance; the output of the multi-dimensional feature recognition model is combined with the disturbance at the prediction moment to obtain the final abnormal passenger flow prediction result.

7. The integrated intelligent operation system for comprehensive transportation hubs based on multi-source heterogeneous data according to claim 1 is characterized in that: The further analysis of the actual real-time passenger flow and average passenger stay time of the hub based on the real-time passenger flow information collected by the multi-source data collection module specifically includes: The real-time passenger flow monitoring module obtains the actual real-time passenger flow of the hub based on the real-time passenger flow information collected by the multi-source data collection module and by combining the mobile phone signaling data of the hub staff: Where T is the total instantaneous passenger flow at the hub. The multi-source data collection module is equipped with n base stations in the hub for collecting real-time passenger flow information. i is the number of people at the i-th base station, ∑C a is the number of staff members of the hub, τ is the conversion coefficient, and the instantaneous total passenger flow in the hub during the historical period is input into the time series model to predict the future short-term passenger flow in the hub; The real-time passenger flow monitoring module also analyzes the average passenger stay time through the following formula: in, represents the residence time of the u-th passenger at the i-th base station, represents the average stay time of N passengers at the hub.

Images